CONFLATION OF THERMAL BRIDGING ASSESSMENT AND BUILDING

THERMAL SIMULATION

*Abdullatif E. Ben-Nakhi, **Essam O. Aasem

* College of Technological Studies, P.O. Box 3665, 22037, Salmiya, Kuwait

** Kuwait Institute for Scientific Research, P.O. Box 24885, 13109 Safat, Kuwait

ABSTRACT
causes:
The legislation and energy awareness have lead to
increased thermal insulation levels in buildings.
Consequently, heat flow, between the indoor and
outdoor environments, due to thermal bridging is
forming an increasing fraction of building thermal
load. Accurate thermal bridging assessment is
becoming more important not only to predict the heat
flow, but also to predict the level of condensation and
mould growth in the heating season. This paper
presents a thermal bridging assessment module that is
integrated within a state-of-the-art, whole building
simulation environment in order to have more
pragmatic boundary conditions.

INTRODUCTION
In general, heat flow through building construction is
one-dimensional (i.e., in the direction perpendicular
to the surface). This is because thermal conductance,
and temperature differential in the perpendicular
direction is much greater than that in the lateral
directions. However, localized multi dimensional
heat conduction through building envelope is
common. Thermal bridge is the part of building
envelope through which heat conduction is multi
dimensional.
The multi dimensional character of heat conduction
affects the local temperature distribution and heat
flow rate. In other words, thermal bridging will bring
the internal surface temperature nearer to the other
side environment, and causes higher heat flow
between the two environments. While the internal
surface temperature should be considered in mould
growth and condensation risk assessments during the
heating season, higher heat flow rate should be taken
into account during the design of buildings and their
year round environmental control systems.
In general, buildings have several thermal bridges,
which occur due to one or more of the following

1- Change in thermal properties of building

envelope in the lateral direction.
2- Change in construction thickness (e.g. a
window within a wall).
3- Difference between internal and external
areas (e.g. edges and corners).
4- Heat
generation
within
construction (hot water pipe).

building

In addition, the legislation and energy awareness

have lead to increased insulation levels in buildings,
which implies increased thermal resistance in the
perpendicular direction.
Consequently, the
difference between thermal conductance in the
perpendicular and lateral directions is reduced, which
increases the potential for multi dimensional heat
conduction in building constructions. Furthermore,
the severity of the thermal bridging due to the causes
mentioned above has increased.
Traditionally, facilities available to designers for
assessing thermal bridging have involved either the
use of guidebooks or general-purpose numerical
analysis tools. The former suffers from the handicap
that the design details in question does not
necessarily match the details in the guidebooks. The
latter can be time consuming to set up, unable to
model multiple dynamic boundary conditions for a
domain, and are not integrated with other heat
transfer processes in the building.
There have been several recent initiatives aimed at
improving this situation. Strachan et al (1995)
introduced two such developments: the KOBRA
software, which allows for 2D thermal bridge
analysis, and the 3D conduction capabilities within
an existing building energy simulation package. The
former combined with the EUROKOBRA database,
consisting of some 3000 thermal bridge details,
provide a user-friendly way to analyse such thermal
bridges. Although this method simplified domain

definition, the problem of boundary conditions was

not tackled.
The later tool facilitates multidimensional modeling of heat conduction through
building construction. The gridding flexibility of this
tool is inherited from the adopted unstructured1
mesh, which has some shortcomings such as high
CPU effort and space.

The control volume formulation is achieved by

integrating the associated partial differential equation
(Eqn. 1) over a small control volume. Accordingly,
for a rectangular parallelepiped control volume, with
heterogeneous material and uniform boundary at
each surface we have
Nm

Accordingly, the current project was invoked in order

to develop an adaptive thermal bridging assessment
tool. This was achieved by incorporating structured
3D dynamic conduction capabilities into an existing
building energy simulation environment.
After
introducing the theoretical background, the
integration of the module into a state-of-the-art
building simulation package is presented. Then, the
numerical and analytical validation of the developed
tool is presented. Finally, conclusions are given.

i ci Vi
i =1

6
T
= As qs + V g
t
s =1

Where the heat flux qs can be due to heat

conduction, convection, or radiation, which are
defined by

qcond =

ji (T j Ti )
x j i

qconv = hc Tj Ti

THEORETICAL BACKGROUND

cp

b


T (r , t )
= qi + g (r , t )
t
i =1

(1)

Many methods for the numerical formulation of the

above equation exist. The control volume approach
was adapted in the present work because of its
physical elegance, and its formulation simplicity and
flexibility. In addition, there are several possible
schemes for the positioning of control volumes and
their associated grid points. The approach adopted in
the current work is based on putting one node on
each material or boundary interface.
Then
positioning additional nodes in between according to
the required resolution.
After that, the control
volume surfaces are located midway between grid
points.
This approach ensures continuity in the
boundary conditions throughout each control volume
surface, and continuity in the thermal conductance
between nodes.

qrad = hr (T j Ti )

Usually, heat conduction occurring in buildings is

transient. That is because of the changing boundary
conditions, which is affected by the outside climate,
plant operation, occupants activity, etc. The
differential equation of heat conduction can be
written as

In unstructured mesh the identification of grid

points should be individually specified and they are
not associated with an orderly defined grid lines.

(2)

(3)

(4)
(5)

Based on the foregoing theory a structured 3Dgridding scheme has been developed. The main
advantage of this scheme is that it enables localized
3D modeling.
For example, a building can be
modeled as a 1D problem except for parts of it that
are represented by 3D model. The integration of the
new scheme within a state-of-the-art simulation
environment in presented in the next section.

IMPLEMENTATION
It is out of the scope of this paper to present the
theory behind the building energy simulation
represented here by the ESP-r (Clarke 1985), which
is well established and reported in great detail
elsewhere. However, a brief description of the ESP-r
environment
is necessary to present the
implementation of the developed scheme within
ESP-r.
The ESP-r is a tool for the transient simulation of
heat and fluid flow within combined building/plant
systems with control imposed. The structure of ESPr is shown in Figure 1. By means of Project Manager,
a simulation problem is defined by a set of data files
whose names and locations are saved in a single
system configuration file. By defining the system
configuration file name to the Simulator, it will
represent the problem by its equivalent network of
time dependent thermal resistances and capacitances
subjected to dynamic potential differences.

USER DOMAIN

TECHNICAL DOMAIN

Flow
Control
Project
Manager

Product
model

Plant

User

Applications,
databases,
tutorials,
etc.

Databases
Simulator

Fabric

Developers

Climate

Shading
Result
Analyzer

Time series
state
variables

Graphics
Design exemplars
3D Gridding

Figure 1: The ESP-r system (Clarke 1994)

By performing a simulation, the Simulator creates a

result file that is analyzed by the Results Analyzer.
Accordingly, three levels of integration between the
developed scheme with the ESP-r environment are
considered. These levels are problem definition,
simulation, and result analysis (Figure 2). Since
ESP-r is equipped with an advanced gridding
module called grd, the required data for the
structured 3D scheme are defined within grd.
However, the 1D problem should be defined within
the ESP-r environment first.
The developed scheme deals with one zone at a
time. Each zone is divided into local components
and one imported domain. The local components
are the default 1D constructions, which can be set
to defined or not defined. The undefined local
components are those components that will be
represented by the imported domain. Therefore, in
order to model 3D-heat flow through, say, the East
and South walls, they should be set to undefined
local components and a 3D model of the two walls
and the edge in between should be defined and
imported.
The imported domain is defined with respect to a
cartesian coordinate system. The definition of the
imported domain requires three sets of data: grid
data, material geometry, and boundary conditions.

The required grid data for the imported domain are

the employed length unit (e.g. mm or cm), number
of gridding lines, and distance between each two
successive grid lines in each dimension. This group
of grid data facilitates high level of gridding
flexibility. The internal and external boundaries are
referenced to existing boundaries in the 1D
problem. Similarly, the thermo-physical properties
of the imported domain are defined by referencing
to existing layers within the 1D building
constructions.
In order to encourage the usage of the developed
scheme, it is furnished with the default ESP-r
interface for the definition of the imported domain.
Beside that, on-line help and exemplar are also
provided.
The simulation of a problem within the Simulator is
performed in a three-stage process: discretisation of
the problem, derivation of the simulation equation
for the nodal system, and simultaneous solution of
the derived characteristic equations. The default
ESP-r space discretisation approach is based on 1D
heat conduction through building constructions.
Accordingly, each inter-constructional node has
two heat conduction connections.
However,
construction surface nodes have only one
conduction connection.
Depending on the
boundary conditions, the other connections for the
construction surface node are defined. For example,

Project Manager

Geometry

Connections

Construction

Configuration

Control

Operation

Plant

Flow

etc.

1-D Model Definition

grd

Local components
indices

Grid data

Material
geometry

Boundary
conditions
Monitored nodes
details

Imported domain
Simulator

Structured Mesh File

Default ESP-r Routines

Discretisation of the
imported domain

Computation of external
and internal solar gains

Discretisation of defined
local components

Computation of external
longwave radiation
Computation of
convective and radiant
casual gains

Computation of internal
longwave radiation

Loop
for
each
time
step

Modification based on
volume or area
weighting

Computation of airflow
thermal load

Setting up the system

matrix
etc.

Direct or iterative
solution of the matrix
Temperature estimation for
internal surface node based
on area weighting

External boundary

Temperature calculation
for external surface and
inter-construction nodes

Saving the plant flux

and nodal temperatures
1D Nodal Temperature Calculation

Text Editor

Result Analyzer

Figure 2. Schematic showing integration of the new module within the ESP-r simulation environment.

the internal surface nodes have one convection

connection with the zone air node, and radiation
connection with other internal surface nodes.

Furthermore, a climate data file defines the building

external boundary variables.

With respect to the employed discretisation, a

system matrix is created. In this matrix each node
(i.e. 1D and 3D) is represented by one equation.
These equations are then solved simultaneously
with respect to the invoked control law. Two
solution methods are available: direct and iterative.
The iterative method is the default one, as it
requires less storage space and it produces less
round off errors. The adopted direct and iterative
solution methods are Gauss's elimination method
and the Gauss-Seidel method, respectively. The
adopted Gauss-Seidel method incorporates linear
under-relaxation factor.
For the defined local components, the default ESP-r
space discretisation will be employed and their
associated characteristic equations are created in the
system matrix. The undefined local components
will not be directly represented in the system
matrix. They are replaced by the imported 3D
domain. As in the 1D gridding, the internal surface
nodes are connected with the space air node by
convection, and connected with each other by
radiation. For the internal radiation calculations,
the 1D-view factors are employed after area
weighting the 1D values and setting to zero the
view factors between nodes within the same
surface.
The fully implicit discretisation scheme is
employed when the 3D modeling is invoked. This
is because the fully implicit scheme is
unconditionally stable, the coefficient generation
process requires less CPU effort compared to other
implicit schemes, and the amplification factor is
always positive, hence prevents oscillation in the
results, as shown in Figure 4.
Based on the results obtained from the 3D scheme,
the default 1D temperatures are estimated either
directly for the 1D constructions or by volume
weighting for 3D constructions. This is because the
1D temperature distribution is required in the
calculation of other thermal processes such as
shortwave radiation, heat absorption by transparent
materials, and convective heat transfer coefficient
estimation.
At the result analyses level, the default options
within the ESP-r environment, such as heat fluxes
and temperature distribution can be used after 3D
simulation. In addition, the new scheme allows
monitoring the temperature profiles for several
predefined nodes within the imported domain.

VALIDATION
In general, validation processes fall largely into
three categories: analytical verification; inter-model
comparison; and empirical validation. In analytic
tests the predictions of programs are compared with
exact analytical solutions. In empirical validation
the results from thermal programs are compared
with measurements made in buildings. In intermodel comparisons the predictions of a program are
compared with those of other programs which,
usually, are of similar sophistication.
The developed scheme was validated by intermodel and analytical verifications only, as no
associated empirical validation data was available.
In inter-model comparison, the results of the
developed 3D module, which was integrated within
the ESP-r, were compared against the ESP-r's
default 1D analysis and VOLTRA packages. First,
the developed module was used to model transient
3D-heat flow through a building construction
similar to the wall construction in Figure 3. The
boundaries at the lateral directions were set to
adiabatic. For the perpendicular direction, the
internal ambient temperature was set to 24 C, and
the external boundary was defined by climate file of
a typical meteorological year for Kuwait.
Therefore, the defined problem was of 1D nature
even though a 3D gridding was employed. The
heat flow rate for the 3D problem were compared
with that of an equivalent 1D model by ESP-r. The
results matched up to two decimal digits.
For the other inter-model comparison, the
developed module within the ESP-r was invoked to
compare against VOLTRA modeling accuracy. The
problem modeled by VOLTRA was transient heat
conduction through the building corner shown in
Figure 3. As shown in Figure 4 and Figure 5 the
results agree with each other except for minor
variations. The minor differences in the results
were due to the difference in the discretisation
schemes employed. While ESP-r is based on fully
implicit scheme, VOLTRA incorporates CrankNicolson discretisation scheme. The oscillation in
the VOLTRA results is due to the nature of stability
error
associated with the Crank-Nicolson
discretisation scheme (Hensen and Nakhi 1994).
For 10 minutes time step, the oscillations were
significantly dampened and better agreement was
obtained between the ESP-r and the VOLTRA. The

VOLTRA is a tool developed by the Belgian

company PHYSIBEL for 3D transient heat
conduction modeling.

Roof
Tile
Cement mortar
Insulation
Reinforced concrete
Cement mortar
Cement Block
Insulation
Cement mortar
Brick
Figure 3: A schematic showing horizontal and vertical cross-sections of a corner

results were not shown since they almost overlap

each other.


)C4 ( erutarepmeT

In the analytical verification, the results of ESP-r

were compared with the exact solution of transient
three-dimensional heat conduction through a
homogeneous
slab.
For a rectangular
parallelepiped domain (0 x a, 0 y b, and 0
z c) that is initially at 50 C, and for times t > 0
the boundaries are defined by



VOLTRA



ESP-r










 

T
+ hout T = 0
x

Figure 4: External corner temperature

(6a)


at x = a

(6b)



)C4 


T
+ hinT = 0

at x = 0



T
=0
y

T
+ h yT = 0
y

T
=0
z

T
+ hz T = 0
z

at y = 0

(6c)



VOLTRA
ESP-r



at y = b

(6d)






 

Figure 5: Internal corner temperature

at x = 0

(6e)

at z = c

(6f)

Ozisik (1993) presented the general solution for

multi-dimensional homogeneous heat conduction
problems. For the current problem the solution can
be written as



T (x , y , z , t ) = 8T * e

in the external ambient temperature was also

intended to amplify the errors.

m2 + n2 + 2p t

50
analytical

ESP-r

4


m cos( m x ) + out sin ( m x )

2
2 hout
hout
hin
m + 2 a +
+
2

2 hin

m + 2

45

40

35
0

2 h y2 2 h z2
cos( n y ) cos( p z ) n + 2 p + 2

2 h y2 h y 2 hz2 hz
b n + + c p + +

2
2

Where, the eigenvalues

m , n , and p are the

m (hout + hin )
h h

m2 out 2 in

n tan ( n b ) =

hy

p tan ( p c ) =

hz

12

16

20

Time (min)
Figure 6: Comparison between analytical and ESP-r
results

CONCLUSIONS

(7)

positive roots of the following equations

tan ( m a ) =

A new thermal bridging assessment module that

was integrated within a state-of-the-art, whole
building simulation environment was developed.
The developed tool is distinguished by the
flexibility in domain definition, and by the level of
conflation with whole building simulation package,
which
facilitated more pragmatic boundary
conditions for the domain under study.

h
hout

+ m sin( m a) out cos( m a )

m n p
sin ( n b ) sin ( p c )

(8)

(9)

(10)

The results of the analytical validation based on 500

eigenvalues of each of , , and are shown in
Figure 6. The temperature profiles shown are for
the center node of the external surface. The reason
for selecting the external surface is because of the
expected highest truncation error due to the highest
temperature gradient in the time direction. In
addition, the stepwise excitation from 50 C to 0 C

In order to encourage the employment of the

developed tool in practice, it was furnished with a
user-friendly interface that is compatible with the
whole building simulation package. In addition,
both inter-model and analytical validations were
performed to verify the adopted domain definition
procedure. The tool was equipped with further userfriendly features such as the on-line help and
exemplars.

REFERENCES
Clarke, J. A., 1985, Energy Simulation in Building
Design, Adam Hilger Ltd., Bristol.
Clarke, J. A., 1994, Building Simulation: Realising
the Potential, Proceedings of the Building
Environmental Performance, U.K.
Hensen, J. L., and A. E. Nakhi, 1994, Fourier and
Biot Numbers and the Accuracy of Conduction
Modeling,
Proceedings
of
the
Building
Environmental Performance, U.K.

Ozisik, M. N., 1993, Heat Conduction, John Wiley,

USA.
Strachan, P. A., A. E., Nakhi, and C., Sanders,
1995, Thermal Bridging Assessments, Proceedings
of the Building Simulation `95, U.S.A.

NOMENCLATURE
A

Area (m2)

cp

Specific heat (kJ/kg.K)

Heat source term (W/m3)

hc

Convective
(W/m2.K)

hr

Radiative heat transfer coefficient (W/m2.K)

Nb

Number
of
conditions.

Nm

Number of homogeneous materials in a

control volume.

Time (s)

Temperature (C).

T*

Initial temperature (C).

Volume (m3)

x, y, z

Location in the x, y, and z directions (m)

Density (kg/m3).

Thermal conductivity (W/m.K).

, , Eigenvalues.

heat

transfer

homogeneous

coefficient

boundary